signal-bot/DOCS-photo-pack.md
Krzysztof kuhy Rudnicki aefffb2fd9
Some checks are pending
CI / check (push) Waiting to run
feat: add a photo to the album from the chat, tagged, stripped and verified
Send a photo captioned "!add_photo alfa wakacje" in one of the album's own
groups and it lands in the pack tagged with both, retrievable through !alfa
and the new !losowe immediately -- no restart, and no tagger session on the PC.
A label matching nobody becomes a custom tag, and !losowe is the only door to
one, which is what keeps a live swap a *data* swap: it can never register a
trigger or widen a command's groups.

No typed label travels. Labels become pseudonym keys in a new registry
arg_resolver, at the dispatcher, before the line that logs the command name --
so no handler, journal entry, command_uses row or reply body ever holds a word
somebody typed. The reply is counts and pseudonyms: "dodano · p1: 4 · t1: 1".

Stripping is exiftool with a *positive* check on the output, never an exit
code. ffmpeg was tried first and fails silently on photos: given a JPEG with a
real APP1 Exif segment it returned a byte-identical file, GPS intact, exit 0.
The surviving-tag allowlist is per tag *name* rather than per group, because
measurement showed a group-level "must be empty" rule rejects every format
except JPEG -- a valid PNG always reports its IHDR, an MP4 some fifty QuickTime
tags that are the container itself. Codec configuration (CompressorName,
VendorID, the HEVC parameter set) is on the list by decision: CompressorName
lives in the avc1 sample description and exiftool cannot write it at all, so
the alternative was refusing every video and every iPhone HEIC. Location and
identity tags are still refused, and the e2e run re-reads the stored files to
prove a QuickTime geotag did not survive.

.webm and .svg stay on the format allowlist and are stored *unstripped*:
exiftool 13.55 can write neither, and its SVG reader does not even surface
<metadata>. That is a deliberate, documented hole, not an oversight.

Two things this had to defuse along the way:
- personal_data_sources now reads custom.json, so "!add_photo kot <photo>"
  would have outlawed the word "kot" and refused every future commit, fixable
  only by hand-editing an album the committer cannot see. Minting now refuses
  a word that is already a command, and the gate subtracts the repo's own
  alias vocabulary from custom labels only -- curated people.json names are
  never excused.
- signal_bot/external.py is now the single subprocess site in the repository,
  so ruff's S603 needs exactly one scoped per-file-ignore instead of one per
  module that shells out. A test asserts nothing else imports subprocess.

A failed add changes nothing: media and the three manifests are written, the
album is reloaded, and only a load that succeeds is published; a load that
fails rolls all four back and keeps serving the pack that was already working,
without raising -- Restart=always would turn a ConfigError into a loop.

977 tests, 100% branch coverage, no omits and no suppressions. scripts/e2e.sh
gains 8 scenarios driven against the real exiftool on media generated at run
time, and asserts the invariants a reply cannot: that a refused upload is not
stored, that a message with no attachment leaves custom.json byte-identical,
and that signal-cli's own copy of every ingested attachment is gone.

Closes TODO-add-photo.md.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JqqE22jp94B1ypwxppdr1o
2026-08-31 11:20:00 +02:00

228 lines
11 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# The photo pack
A pack is a private album the bot posts from: `!alfa` returns a random photo
of alfa, and `!alfa beta` returns one with both of them on it.
**Nothing about a real pack is in this repository, and nothing ever will be.**
Who the people are, what somebody types to get their photo and which file
shows whom live in one directory outside the repo. This document, the tests
and the offline end-to-end run all use the same invented people — alfa, beta,
gamma — so the mechanism can be public while the album stays private.
## Where it lives
```
~/.local/share/signal-bot/pack/
people.json # the commands, and the only place names exist
tags.json # which file shows whom
media/ # the photos and videos themselves
```
Found in this order: `PHOTO_PACK_DIR`, then `$STATE_DIRECTORY/pack` (what
systemd hands a **system** unit), then `$XDG_DATA_HOME/signal-bot/pack`.
`StateDirectory=` is deliberately **not** in the unit this repo installs:
for a *user* unit systemd resolves it below `$XDG_STATE_HOME`, which is not
where the album is. Discovery finds the same directory, and leaving
`PHOTO_PACK_DIR` unset is what keeps a not-yet-filled-in pack from turning
`Restart=always` into a restart loop.
Two absences, deliberately different:
- a pack **nobody named** that is not there, or that has no `people.json`
yet, is an installation without an album: the commands do not exist, and
that is the normal case on every machine and in the whole test suite;
- a pack named in `PHOTO_PACK_DIR` that is missing, or whose manifest cannot
be honoured, **fails startup**. `load_config` raising is an answer, and a
bot that came up healthy and said "no such photo" forever would not be.
## people.json
```json
{
"groups": ["group.WlRKbExYZGhkR05vWldRdA=="],
"people": [
{"key": "p1", "triggers": ["!alfa", "!alfka"]},
{"key": "p2", "triggers": ["!beta"]}
],
"combos": [
{"key": "c1", "triggers": ["!oboje"], "all_of": ["p1", "p2"]},
{"key": "c2", "triggers": ["!ktokolwiek"], "any_of": ["p1", "p2"]}
]
}
```
- **`groups` is the whole privacy guarantee.** In any other group these are
not commands at all: not answered, not counted as a command, not named in a
log line. A handler that quietly returned would still leak all three, which
is why the *dispatcher* enforces it — `registry.resolve_with_args` takes the
group and gives back nothing outside it.
- **`key` is a pseudonym**, and the only identifier that reaches the journal
or the `command_uses` table. A command is `pack:p1`; reading either says
somebody asked for a photo and nothing about whose.
- **`all_of` means one photo with all of them; `any_of` means either**, and
where a choice exists the bot picks among the readings that actually have a
photo, so a combination that could be answered is not refused.
- Triggers may not contain a digit — `folding.py` maps `0134578` onto letters,
so `!alfa7` is not what you typed. Two spellings that fold the same are fine
*within one entry* (`!celina`/`!celiną` is one person written twice) and
refused across two.
## tags.json
```json
{"alfa.png": ["p1"], "both.png": ["p1", "p2"], "clip.mp4": ["p1"]}
```
A tag naming a file that is not in `media/` fails startup: a renamed file
would otherwise drop out of the album in silence. An empty list means "seen,
nobody in it", which is how the tagger records a photo you skipped.
Videos work. The attachment goes out as
`data:video/mp4;filename=<name>;base64,<data>` — the spelling is the
container's own, from its `/swagger/doc.json` — because told it is a photo,
Signal renders a clip as a file nobody can play.
## custom.json
```json
{"t1": ["wakacje2019"], "t2": ["gorybieszczady"]}
```
Labels `!add_photo` learned from the chat that match no person. The key is a
pseudonym in the same family as `p1` — it is what reaches `tags.json`, a log
line and a reply — and the label is the only place the typed word is kept.
Custom tags define **no commands**: `!random <label>` is the only door to one.
That is deliberate, and it is what makes a live album swap safe — a swap
replaces pack *data* only, and can never register a trigger or widen a
command's groups.
A label that folds onto a person's trigger, or onto another key's label, fails
startup rather than shadowing the command it collides with.
## adds.json
```json
{"a3ae52c7….jpg": {"by": "<uuid>", "at": 1756600000}}
```
Who added which file. **Nothing reads it** — `pack_load` ignores the file
entirely — and it exists only so `!undo_photo` can be built later without a
gap in its history; the adder has to be recorded at add time or not at all.
It lives inside the album, which is `0700` and outside this repository, so it
is handled exactly like the photos.
## Adding from the chat: !add_photo
Send a photo captioned `!add_photo alfa wakacje2019` in one of the album's
groups. `alfa` resolves to that person's key; `wakacje2019` matches nobody and
is minted as a custom tag. The photo is retrievable through `!alfa` and
`!random wakacje2019` **immediately**, with no restart.
What holds:
- **No typed label travels.** Labels become keys in the registry's
`arg_resolver`, at the dispatcher, *before* the line that logs the command
name. The reply is counts and pseudonyms — `dodano · p1: 4 · t1: 1`.
- **No cap.** No cooldown, no size limit, no daily limit. `!random` takes the
ordinary pack cooldown; `!add_photo` takes no decorator at all.
- **At most six labels**, the same ceiling `!alfa beta …` has.
- **The attachment must be on the same message**, never a quoted one: a reply
carries the *quoted* message's media in `quote.attachments`, and taking it
would let anyone re-file somebody else's photo.
- **A duplicate changes nothing.** The stored name is a content hash, so a
photo already in the album is reported and skipped; a mixed batch stores the
new files and says how many were already there.
- **Nothing is written until it works.** Media, `tags.json`, `custom.json` and
`adds.json` are written, the album is reloaded, and only a load that
succeeds is published. A load that fails rolls all four back and keeps
serving the album that was already working.
- **The second copy goes.** signal-cli keeps its own copy of every inbound
attachment outside the album's `0700`. `!add_photo` deletes it once the file
is stored, and also when the upload was refused.
### Stripping: exiftool, with a positive check
`ffmpeg -map_metadata -1 -c copy` was tried first and **fails silently on
photos**: given a JPEG with a real APP1 Exif segment it returned a
byte-identical file, GPS marker intact, and exited 0. `-c copy` passes the
compressed frame through as an opaque packet and a JPEG's Exif lives inside
it. Do not reintroduce ffmpeg for this.
So the gate is a **positive check on the output**, never an exit code:
```sh
exiftool -all= -o <dst> <src> # never --overwrite_original on the temp
exiftool -G0 -s -all <dst> # then read it back
```
and every surviving tag must be **structural**. Measured against exiftool
13.55, the plan's original "the surviving set must be empty" would have
rejected every format except JPEG — a valid PNG always reports its IHDR under
`PNG`, a GIF its screen descriptor under `GIF`, an MP4 some fifty `QuickTime`
tags that *are* the container. So the allowlist is per **tag name**
(`signal_bot/media_tags.py`), which keeps the property that matters: an atom
nobody anticipated arrives as a name that is not on the list, and the upload
is refused.
Allowlist, not blocklist, for the reason a naive gate misses: an MP4 geotag is
a QuickTime `udta`/`©xyz` atom reported under **`QuickTime`**, not `GPS`.
The list includes the **codec's own configuration** — `CompressorName`,
`VendorID`, and the HEVC parameter set (`HEVCConfigurationVersion`,
`GeneralProfileIDC`, `BitDepthLuma`, …). That was forced by measurement, not
chosen for convenience: with them off the list exiftool 13.55 could produce no
acceptable video *and* no acceptable HEIC, because `CompressorName` lives in
the `avc1` sample description and exiftool reports it as not writable at all
(`Sorry, QuickTime:CompressorName doesn't exist or isn't writable`), and a real
HEIC reports sixteen HEVC config tags. Since HEIC is the iPhone default, the
choice was codec parameters or no phone photos. They describe how the picture
is encoded, never where it was taken or who took it; every location and
identity tag is still refused, and `scripts/e2e.sh` proves a QuickTime GPS atom
does not survive on a file the album actually stored.
Formats are decided by **magic bytes**, never by the filename: jpg, png, gif,
webp, mp4, mov, m4v, heic, heif, avif, webm and svg. Every one of them must
also be in `pack.MEDIA_TYPES`, or a file that uploaded cleanly is refused by
the *next* load and takes startup down.
### ⚠️ .webm and .svg are stored unstripped
exiftool 13.55 can write neither (`exiftool -listwf` lists no `WEBM` and no
`SVG`; a strip attempt returns `Can't write SVG files`). Its SVG *reader* does
not surface `<metadata>` content either, so there is no detection to fall back
on.
kuhy's decision of 2026-08-31 is to keep both on the allowlist and store them
**as they arrive, metadata included**. The consequence, stated plainly: an
`.svg` carrying `<metadata>` or an embedded
`<image xlink:href="data:image/jpeg;base64,…">` keeps whatever is in it, and
every later `!alfa` re-sends that file verbatim. Every other format on the
list is stripped and verified.
## Filling it in: tool/tagger.py
```sh
python tool/tagger.py --import ~/photos # copy the files in, then tag
python tool/tagger.py # carry on where you left off
```
It serves one page on `127.0.0.1:8140`: the item, a numbered button per
person, `1`–`9` to toggle, `→`/`←` to move. Every toggle rewrites `tags.json`
atomically, so closing the tab loses nothing. Names are typed **into the
page**, never as arguments — an argument would land in the shell history —
and the group is picked from the live `/v1/groups` listing rather than pasted.
When you stop it, it prints how many items were never opened and how many
photos each command would have. **A command showing `0` is dead on arrival**;
that is the number to check before restarting the bot, not after somebody in
the group tries it.
## Running the mechanism without an album
`scripts/e2e.sh` writes a synthetic pack with `tool/e2e_pack.py` and proves,
offline and with no Signal account, that one name posts a photo, two names
intersect, an impossible combination gets one line and no attachment, a video
leaves as a video, and the same command in another group produces **zero**
sends. It also drives `!add_photo` against the real exiftool, on media it
generates at run time, so no photo enters this repository.